Terminus

Terminal-Bench is a benchmark for testing AI agents in real terminal environments, evaluating how well agents can handle real-world, end-to-end tasks autonomously. The benchmark includes tasks spanning coding, system administration, security, data science, model training, file operations, version control, and web development. Terminus is the neutral test-bed agent designed to work with Terminal-Bench, operating purely through tmux sessions without dedicated tools.

Released

Unknown

Readiness

Paper only

Primary field

General AI

Benchmark Radar records only publicly supported details and links back to primary sources for verification.